Comparison of pareto fronts for pump impeller design using sobol sequence sampling with water and methanol

This study focuses on the optimization and performance evaluation of pump impellers for water and methanol using Sobol sequence sampling, Artificial Neural Network (ANN)-based metamodeling, and Multi-Objective Genetic Algorithm (MOGA) optimization. Initially, 40 design points generated via Sobol sequences facilitate the exploration of a multidimensional design space, enabling the design of impellers with varied geometrical parameters. The resulting head and efficiency values are used to train an ANN model, achieving high accuracy, with overall R-values above 0.99 for both fluids. Optimized impellers for water and methanol show improved flow uniformity and energy efficiency, as evidenced by smoother velocity distributions. For water, the optimized impeller achieved a head of 10.01 m and an efficiency of 72.41 %, while for methanol, it reached a head of 10.01 m and an efficiency of 73.62 %, as obtained by CFD. Pareto analysis reveals that water designs are constrained around a 10 m head, whereas methanol allows flexibility, achieving optimal efficiency across a 10–15 m head range. These findings confirm the efficacy of the optimization framework, offering an adaptable approach for enhancing pump impeller performance across different fluid applications.